Network Optimization of Optical Performance Monitoring Lian K. Chen ( 陳亮光 ) The Chinese University of Hong Kong Hong Kong, SAR, China The Chinese University of Hong Kong 1
Outline Motivations Advanced Monitoring Techniques for point-to-point link OSNR Monitoring Chromatic Dispersion Monitoring Multi-impairment Monitoring Optimization of performance monitoring networks Optimization of Fault-diagnosis with Minimum Probing Algebraic Monitoring Network Other related works Conclusions The Chinese University of Hong Kong 2
Optical Performance Monitoring Definition: Physical layer monitoring of the signal quality, for the purpose of determining the health of the signal in the optical domain OPM OPM OPM OPM Ref: A. Vukovic, H. Hua, M. Savoie, Performance Monitoring of Long-haul Networks, Lightwave Magazine (2002) Applications: Signal quality characterization Fault management Active compensation Quality of service (QoS) provisioning Examples: Relating OSNR with BER Early signal degradation alarm Fault detection, localization, and isolation Resilience mechanism activation Dynamic CD + PMD monitoring and compensation SLA fulfillment verification The Chinese University of Hong Kong 3
Wish List for the Enhancements of OPM higher dynamic range; higher sensitivity; robustness to various effects; multi-impairment monitoring; and lower cost. But how? Optimization of Performance Monitoring System The Chinese University of Hong Kong 4
OSNR monitoring module PM driven by Plow sig, in NEB frequency f OSNR( db /0.1 nm) = sinusoidal signal PASE, in 0.1nm CW & CCW (1 r) experience NEB periodic f = phase difference [ r(1 Dsig ) (1 + Dsig )] 0.1nm Signal switched out periodically P max Track /P P max (Pand max -PP min min )/(P max to +P calculate min ) OSNR min when noise is absent bandpass polarization filter isolator scrambler Phase Mod Proposed OSNR Monitoring Module (Ref: Y.C. Ku et al., OFC 2006) polarization controller Power Meter Transfer function (db) λ signal λ Output power P max P min V Phase Mod. The Chinese University of Hong Kong 5
Experimental results: 10Gb/s, back-to-back Without PMD: Monitoring error < 0.25 db 40-dB monitoring dynamic range With PMD: Influence of DGD reference OSNR 25 db/0.1nm Monitoring error < 0.25 db OSNR by proposed scheme (db) 50 40 30 20 10 0 4 2 0-2 -4 0 10 20 30 40 50 Monitoring Error (db) Monitoring errror (db) 4 2 0-2 -4 0 10 20 30 40 50 OSNR by OSA (db) DGD (ps) The Chinese University of Hong Kong 6
Experimental results: 10Gb/s, 150km, w. ~1.5ps PMD Influence of chromatic dispersion and transmitted power An EDFA was added after each 50-km SMF span EDFA output power 1 or 10 dbm OSNR (db/0.1nm) 30 28 26 24 22 20 OSNR by OSA, P out = 1dBm OSNR by proposed method, P out = 1dBm OSNR by OSA, P out = 10dBm OSNR by proposed method, P out =10dBm 0 50 100 150 Distance (km) The Chinese University of Hong Kong 7
Experimental results: 10Gb/s, back-to-back, w/ polarized noise Influence of partially polarized noise Worst case: DOP=99.69% OSNR by proposed scheme (db) 50 40 30 20 10 0 0 10 20 30 40 50 OSNR by OSA (db) ASE noise // Signal ASE noise // Signal ASE noise _ _ Signal ASE noise _ _ Signal 4 2 0-2 -4 Monitoring Error (db) The Chinese University of Hong Kong 8
Chromatic Dispersion (CD) Monitoring The Chinese University of Hong Kong 9
Previous work on chromatic dispersion monitoring Clock phase difference monitoring Pros: No need to modify transmitter, highly sensitive Cons: requires high-speed electronics Ref.: Q. Yu, et. al., IEEE JLT, 20, 2267-2271 (2002) We propose to use a birefringent fiber loop (BFL) to realize a simple, polarization insensitive CD monitoring scheme without transmitter side modification, and demonstrate in NRZ system Ref.: Y.C. Ku, et. al., OFC 2006 The Chinese University of Hong Kong 10
Proposed Dispersion Monitoring Module based on BFL Measure RF power at dispersion dependent frequency dip (f RF ) produced by birefringent fiber loop (BFL) Hi-Bi Fiber Dispersion Monitoring Module 10Gb/s 10 23-1 PRBS SMF DCF PC 1550nm IM EDFA BPF EDFA 3-dB coupler RF Spectrum Analyzer P 2 2 2 ϕω + ϕ Ω 2 πλ f RF sin = sin D 2 2 f =Ω /2π = 1/2τ RF Power (dbm) -42-44 -46-48 -50-52 0.8nm Ref.: Y.C. Ku, et. al., OFC 2006-54 -56 1549.8 1549.9 1550.0 1550.1 W avelength (nm) The Chinese University of Hong Kong 11
Experimental Result -22 450ps -24 110ps Power at 5GHz (dbm) -26-28 -30 0ps -32-34 -1000-500 0 500 1000 1500 2000 Dispersion (ps/nm) Measured unambiguous monitoring range: 1500 ps/nm for 10-Gb/s NRZ signal In principle, monitoring range can reach c/2λ 2 f RF2 = 2497ps/nm The Chinese University of Hong Kong 12
Multi-impairment Monitoring The Chinese University of Hong Kong 13
Multi-impariment monitoring - Delay Tap Asynchronous Waveform Sampling Combines asynchronous sampling with two tap delay lines, so that each sample point comprises two measurements (x and y), separated by a fixed time corresponding to the delay length. CPU Low Speed A/D Δt Δt=bit period T s T s x 1 y 1 Δt x 3 y 3 y x 2 y 2 Sarah D. Dods and Trevor B. Anderson, Optical Performance Monitoring Technique Using Delay Tap Asynchronous Waveform Sampling, OFC OThP5, 2006 x The Chinese University of Hong Kong 14
Application of asynchronous sampling in PMD monitoring 0ps 10ps 20ps 30ps 40ps 50ps 60ps 70ps 80ps 90ps m: proportion of sample points falled on left and bottom edges 1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0 20 40 60 80 DGD (ps) m = number of sample points falling on the left and the bottom edges total number of sample points on the scatter plot The proportion of sample points on the left and bottom edge is used as the effective parameters to evaluate PMD value The Chinese University of Hong Kong 15
Application of asynchronous sampling in misalignment monitoring for RZ-OOK -50ps -40ps -30ps 0ps +30ps +40ps +50ps The increasing dispersion and rotation of the sample points on the diagonal can serve as effective parameters to evaluate alignment status 1 2 2 π d = xi + yi sin arctan n i 4 1 y t = arctan i n x i i y x i i Ref.: Y.C. Ku, et. al., ECOC 2006 d: deviation from diagonal line t: average angle of sample points The Chinese University of Hong Kong 16
Experimental results d: deviation from diagonal (a.u.) 1.8e-3 0.86 1.6e-3 0.84 1.4e-3 0.82 0.80 1.2e-3 0.78 1.0e-3 0.76 8.0e-4 0.74 6.0e-4 0.72-50 -40-30 -20-10 0 10 20 30 40 50 Timing Misalignment (ps) t: average angle of sample points (rad) Sign and amount of misalignment can both be estimated The Chinese University of Hong Kong 17
Correlation of Monitored Parameters Monitored parameters may be correlated. The monitoring schemes need to be able to differentiate different possible impairments Ref: ITU-T G.697 Optical monitoring for DWDM systems Question: Can we derive one parameter based on the measurement of the other parameters? How about other dimensions of correlations, e.g. spatial & temporal? The Chinese University of Hong Kong 18
Optimization of Performance Monitoring System Consider a performance monitoring system S PM that is formed by all the monitoring devices in a network and the corresponding measurements Optimization of S PM : 1. the minimum number of monitors required CAPEX 2. the optimal locations of the monitors monitor placement problem 3. the minimum number of measurements needed OPEX Issues: static network vs. dynamic (reconfigurable) network distributed vs. centralized monitoring The Chinese University of Hong Kong 19
Centralized vs. Distributed OPM Monitoring Diagnosis distributed monitoring/ centralized diagnosis centralized monitoring/ centralized diagnosis single monitoring point for the whole network The Chinese University of Hong Kong 20
Number and Placement of OPM Optimization of Placement and Amount of OPM needed One-monitor-per-link/component is not optimum if we consider failure probability of link/component scalability with the network size amount of redundant alarm 1. Channel-Based: Design Based on Established Lightpaths E.g. 1 LSP2 4 LSP1 2 LSP3 5 3 Alarm matrix computation 1 4 2 M 5 M 3 M 6 7 Ref: Sava Stanic et al, Efficient alarm management in optical networks, in Proc. DARPA Information Survivability Conference and Exposition 2003, pp. 252-260 8 6 7 Optimal monitor placement M : monitor 8 In-service channel performance based Reduce monitor number considerably For static network only The Chinese University of Hong Kong 21
Number and Placement of OPM 2. Network-Based: Design Based on Network Topology E.g. Break down a network into a number of monitoring cycle and assign a monitor to each, using an extra supervisory wavelength to probe the health of the cycle 1 5 8 1 5 M 8 2 4 6 10 2 4 6 10 3 Ref: Hongqing Zeng et al, Monitoring cycles for fault detection in meshed all-optical networks, in Proc. ICPPW 04, pp. 434-439 7 9 M Cycle finding algorithm 3 7 M 9 M Suitable for dynamic networks Reduce monitor number considerably Does not measure in-service channel quality Extra wavelength needed The Chinese University of Hong Kong 22
Scalable Network Diagnosis for All-Optical Networks via Group Testing Over Graphs X IP OXC a X IP OXC b X 1 X a X b X 6 X 7 X 8 X 2 IP OXC X f The Minimum Number of Measurements Needed X IP OXC c X f X 9 X c IP X OXC e Analogy: fake coins problem IP X d X 5 OXC Objective: to minimize the number of X measurements to identify the fake 4 coin(s) that are of less weight Group Test: r = T(X a,x b,x e,x 1,X 8 ) Y. G. Wen, V.W. S. Chan and L. Z. Zheng, "Efficient Fault Diagnosis Algorithms for All-Optical WDM Networks with Probabilistic Link Failures (Invited Paper)," Journal of Lightwave. Technology, vol. 23, no. 10, October 2005, pp.3358-3371. X e X d X 3 The Chinese University of Hong Kong 23
Asymptotically Optimal Run-Length Probing Schemes via an Information Theoretic Approach 6 1 9 7 8 2 3 Fault Detection 00100000000001000000 F S F S 5 0 4 Fault Localization S S F S F F Probe Syndrome FFSSSFSFFS 1100010110 Guidelines for Fault Diagnosis Algorithms The total number of probing is approximately equal to the entropy of the network state Each probe should provide approximately 1 bit of state information The Chinese University of Hong Kong 24
Network Diagnosis an algebraic approach Objective: To find an efficient way to determine the quality of different paths, so that it can be used as a metric for path computation in the network layer for channel setup For example: Network diagnosis that explores the spatial correlation to minimize the number of monitors Two existing monitoring channels c 1 and c 2 New request to setup c 3 NO estimation S.T. Ho, L.K. Chen, C.K. Chan, On Requirements of Number and Placement of Optical Monitoring Modules in All-Optical Networks, OECC, paper 7A2-2 (2005) The Chinese University of Hong Kong 25
Linear expressions of optical impairments Noise Figure (NF) NF NF NF NF = NF + NF + NF + L+ NF overall 2 3 k = NF 1+ 2 + 3 + L+ N Gc2 Gc3 Gck NF i : Noise figure of the i th hop G ci : cumulative gain up to i th hop Chromatic dispersion (CD) CD = Δλ ( L D + L D + L+ L D ) overall 1 1 2 2 k k Polarization-mode-dispersion (PMD) PMD = PMD + PMD 2 2 2 overall fiber component Δλ : spectral width D: dispersion parameter L i : total length of i th hop Overall impairment for all paths can be derived from impairments of individual fiber link and component The Chinese University of Hong Kong 26
A sample network (The monitoring module contain both monitoring source and monitoring detector.) The Chinese University of Hong Kong 27
Problem formulation M: Monitoring Matrix x: individual impairment y: Accumulated impairment y m L L L m x 1 11 1N 1 y 0 0 0 1 0 0 0 0 0 2 x M O M 2 0 0 1 0 0 0 0 0 0 y=mx y 3 = Μ = M O x=m M x -1 3 M 0 M1 1 0 0O 0 M0 1M 0 y N m N1 L L ML m NN x N y For the previous example, total 9 impairment variables to be determined and 2 monitoring modules are used 9:2 Q: What is the minimum number of monitoring modules and where should they be placed? Assumption: monitoring modules are able to monitor different probing channels from all ports by time division manner or by proper labeling The Chinese University of Hong Kong 28
Bus & Ring networks The Chinese University of Hong Kong 29
Two-link-connected network Impairment ( c ) = Impairment ( c ) Impairment ( p ) x green red The Chinese University of Hong Kong 30
Summary of Algebraic Network Diagnosis Proposed a novel algebraic approach of monitoring optical impairments in both un directional nodes and links Proved that two monitoring modules are necessary and sufficient to support our monitoring scheme in two link connected network For every node that becomes a leaf node after removal of all leave nodes in the original network, installing one monitoring module at it is necessary and sufficient to support the proposed probing scheme For general networks with bridges, first transform the network to a treelike structure and then solve it by divide-and-conquer method With the probing scheme, any single-fault can be located in the whole network The Chinese University of Hong Kong 31
Intelligent Fault Diagnosis for Optical Networks: Detection, Isolation and Prognosis Fault Detection Fault Isolation Correction & Restoration Network Network Data Collection Fault Prognostics Operations & Maintenance Two fault-diagnosis formulation approaches: model-based data-driven Sensor Readings Sensor Readings Model Difference/ Residues Knowledge & Training Inference & Estimate Decision & Classification Decision & Classification YG Wen, MIT Model-based Approach Data-driven Approach The Chinese University of Hong Kong 32
Conclusions OPM in next-generation high-speed transparent reconfigurable networks is essential Enhancements of OPM, including higher dynamic range, higher sensitivity, robustness to various effects, and multi-impairment monitoring, are desirable to make the OPM system more effective. By exploring the correlation of different spatial or time domain information, it is possible to further enhance the efficiency of OPM. Some works on the Optimization of OPM network to derive theoretical bounds for the min. number of monitoring module or the number of probing are presented. The Chinese University of Hong Kong 33
Acknowledgement: Projects supported in part by Hong Kong SAR Government CERG Grant 411006 Contributors S.T. Ho, Z.C. Xie,, Y.C. Ku, C.K Chan, and Y.G. Wen The End The Chinese University of Hong Kong 34